RPA Automation Fleet
96+ hours a month, and a full business day off the close.
- 0Hours per month removed
- 0Business day off the close
- 0Industrial clients reported on
- 0Bed acute care hospital
The problem
A hospital finance department doing reconciliation, reporting and validation by hand, every month, under a close deadline — with the usual result that the interesting analytical work got whatever time was left after the retyping was done.
Why it mattered
Twelve years in one finance department is long enough to stop seeing the process and start seeing the waste. Every month the same people rebuilt the same work, and every month the close absorbed it as though it were a fixed cost of nature.
The old process
- Manual reconciliation across GL and hospital financial systems
- Reports rebuilt by hand each period
- Validation performed by eye, at volume, under time pressure
- Exceptions discovered late, when there was least time to handle them
- Close length treated as fixed
The idea
Redesign the process first, then automate it. Automating a bad process faithfully reproduces the bad process, just quicker and with less opportunity for a human to notice.
The system
A fleet of bots across reconciliation, reporting and validation workflows, integrated with the GL and hospital financial systems — supported by batching logic and process redesign underneath, plus dashboards, variance and productivity models, and GL validation logic used across the organization.
How it works
- Process redesigned and batching logic established before any automation was written
- Bots execute reconciliation, reporting and validation deterministically on schedule
- Integration with GL and hospital financial systems removes manual re-keying
- Exception handling is explicit — failures are surfaced, not silently skipped
- Execution is logged, making the work reviewable rather than merely faster
Before and after
BEFORE
Rebuilt by hand, monthly
- Manual reconciliation
- Reports rebuilt each period
- Validation by eye
- Exceptions found late
- Close length fixed
AFTER
Redesigned, then automated
- Scheduled bot execution
- Reports generated
- Validation by rule
- Exceptions surfaced early
- One day shorter
Design decisions
- Redesign before automation
- I did not simply automate the old process. I redesigned it first, which is where most of the day actually came from.
- Deterministic and logged
- Audit reliability improved specifically because execution became consistent and reviewable, not just because it became faster.
- Explicit exceptions
- A bot that quietly skips a failure is worse than a person who misses one, because nobody is looking.
Controls
- Deterministic, logged execution across all automated workflows
- Explicit exception handling with surfaced failures
- GL validation logic applied consistently
- Reconciliation automation with a reviewable trail
Testing
Run in parallel against the manual process until output matched, then cut over. Known-answer testing against periods already closed and audited.
Result
Removed 96+ hours of manual work per month and took a full business day off that facility's month-end close, while improving audit reliability through deterministic, logged execution.
Impact
- 96+ hours per month of manual work eliminated at the facility
- One full business day removed from month-end close
- Improved audit reliability through consistent, logged execution
- Enterprise dashboards, variance and productivity models used by executive leadership
- Reporting owned for 950+ industrial clients: pricing, AR analysis, revenue validation, contract billing
The 96+ hours and the day off the close are facility-level results at Clinch Valley Medical Center. Stated plainly because a number that travels without its scope is how résumés stop being credible.
What I took from it
The day did not come from the bots. It came from redesigning the sequence and batching the work correctly — the automation just made the new design cheap to run. People consistently attribute the result to the technology, which is why they then buy technology and do not get the result.
What I would build next
The pattern was rebuilt at corporate scale. See file JEA-001.